IUIIL bUVUI IIIu~Lrallull The Great Nebula in Orion photographed through a telescope. The soft greenish hue of the ionized gas gradually dimming toward the edge of the field, with its immobile shadings smooth and mellow in spots, hard and sharp elsewhere, together with the diamond-like scintillations of the four closely packed stars of the Trapezium, present a striking picture. Scientific Services. In0.. Commercial Trust Building Fifteenth and Market Streets philadelphia 2, Fennsylvania Locust 3-3277

The

Computer System

Page 2 Introduction 4 Input

6 Output

8 Layout of data

12 Data transfers 14 How Orion uses time 16 How Orion decides whaf to do next

18 The form of instructions 20 Advantages of using Orion

NOVEMBER 1960 Printed in Enntand PI H LIST DC. 4lA @ Ltd. 1960 I Introduclrrg the

FERRANTI ORION

compueer sysfdm

The ORION system enables a business to introduce automation of its procedures by stages, until eventually every aspect of its opera- tions has been brought under an integrated system of control. The principal features of ORION which enable it to be used in this way are: An extremely fast magnetic tape system, operating at 90,000 characters per second. With this tape system large files of business data can be processed at the speeds required. A complete time-sharing system enabling the to make the best use of a number of input and output devices operating at widely differing speeds with time-sharing card-to-tape and tape-to- card conversion and tape-printing can be carried out economically without recourse to additional off-line equipment. The time-sharing incorporates full program protection, preventing mutual interference of programs under any circumstance. Magnetic core storage extendable from 4,096 words to 16,384 words is available. Additional storage or peripheral equipment may be fitted at any time after installation. Direct transference of data from peripheral equipment to any part of the core store is provided, thus virtually eliminating costly and clumsy buffer storage. Special facilities are included for handling all types of commercial data from all currently available media, for printing results at high speed, and for dealing with the resulting paper. *A mass printer uses xerography to produce

output at prin fing-press speeds

**A fransacfor is a hand-fed card reader

which can sense a pencil-marked card,

interrogafe a compufer, and produce an

edge-punched reply Punched Cards to latest available Punched Card readers (Tu* * of a// types.

.) to remote Ferranti Transactors Marked Cards 1 4. .. Wfh edge-punched reply.

to Ferranfi TR7, TR5 7-channel readers ) at 1000 and 300 characters per second.

Creed Tape * to Ferranti TR7, TR5 5-channel readers b at 1000 and 300 characters per second.

Magnetic Tape to Ampex FR 300 tape decks, ') operating at an instantaneous transfer rate of 90,000 characters per second. For all types of data arising on punched cards, punched in any way, according to any coding system, with provision for interstage reading, alphabetical punching and all types of over- punching.

For interrogating the computer from an out-station, a sales office or a factory floor, by inserting a marked card into the transactor and obtaining an edge-punched reply.

For Sales Order, Job Card, Goods Received data arising as a by-product of Flexowriter document preparation, and for preparing programs. Full range of alphanumeric upper and lower case characters with parity check.

For data arising as a by-product of keyboard accounting machine operations such as ledger-posting or pre-listing, and for information received by Telex.

For input of file data in the extremely short time of 89 microseconds per word - equivalent to about 45,000 punched cards per minute. The magnetic tape may be read in either direction. OUTPUT

Sales Costs Profits om Line Printers of the fastest available punched-card types, computer driven. ' L' 111 ".. .a -.,. ,,a, 8 0 , a , ,

from Rank Xeronic Printer af 3000 lines per minute. Mass Printing Output is taken up on rolls as in a printing press, and I subsequently cut and stacked as required. Stacked Mass Printing

from latest available punched card punches of all types.

from tape control fo Ampex FR 300 Tape Decks at an instantaneous transfer rate of 90,000 characters per second.

from Creed 3000 punch at 300 characters per second. .. 'I For medium volumes of results presented in the same form as conventional punched card tabulations, with very flexible layout, full range of upper case alphabetical characters, and i multiple carbon copies. Such tabulations will be used for internal factory and office control. I I For large volumes of results the Rank Xeronic printer provides the equivalent of a high speed printing press under computer control, with complete flexibility of layout and a wide range of typescript, including upper and lower case lettering in a variety of founts. Printing at this speed is needed for large scale despatching, invoicing, and purchase ordering operations.

For information for the office or shop floor, interpreted punched cards provide a convenient unit document.

I

For carry forward file data for re-input next time, the iYe is processed at 89 microseconds per word, equivalent to about 45,000 punched card equivalents per minute.

I

For printing off-line on Creed or Flexowriter page printers with sprocket-fed stationery, and for Telex transmission. ~

w , .._ I For monitoring information put out by the Organisation Programme during operation. Layout of data

The store of ORION holds words of 48 binary digits representing either alphanumeric words of eight 6-binary-digit characters, or numbers equivalent to 14 decimal-digit integers. Special facilities are included in ORION for rapid conversion of sterling, decimal, or other radix numbers in character form, to the equivalent binary numbers, and vice versa. These facilities enable the whole contents of a punched card, read at the highest available speeds, to be converted in less than 10 % of the card reading time. Twelve tracks are used for recording data on 1-inch Ampex tape which may be written forwards, or read backwards or forwards at an instantaneous rate of 90,000 characters per second, i.e. 89 micro- seconds per word. Blocks on tape are of arbitrary length and will normally be long compared with record lengths. In a typical case, 250-word blocks might be used for records of average length 10 words. These records could be read or written at 45,000 records per minute, corresponding to 33 milliseconds per block made up as follows :- 25 records x 10 words x 89 microseconds=23 milliseconds stoplstart time for block =10 milliseconds 33 milliseconds An ORION word can store any of the following:

8 alphanumeric characters 8 groups of 6 binary digits

FERRANTI .a,.. decimal number up fo =247- 1 as a 48-digit binary number 140737488355327 ......

Floafing-point number up to &bit 40-bit signed fraction exponent about ...... I......

=247-1 pence as a 48-digit binary number .'.-..."...."....,......

or several such items packed in one word product codelquantity in tons/YaCW ie? f&d..n...... I .I ...... Layout on Magnetic Tape one word

Write forwards ,Read backwards 0. forwards 4....., Layout of 25 records of average length 10 words on a 250 word block record I record 2 ...... -record 25 check sum1 end-of-block marker Representation of dafa on a punched card. disciplined punching undisciplined punching I I

* - 111111 111111 Y 6 -Letter F as &bit character - Undisciplined column as two Cbit characters : The FerrantC ORBON computer system I

When Orion uses a peripheral device for the input or output of data the transfer proceeds, and meanwhile the computer is able to con- tinue with other work until the peripheral device requires access to a word in the core store. Then the computer "hesitates" for 16 microseconds (maybe half way through an instruction) to allow the word to be transferred. The computer then continues with its work from where it had left off. The diagram shows simultaneous reading from one tape deck, writing on another, and operation of an (inde- pendent) program. Thus, even when transferring to and from magnetic tape, more than 80 %of the full speed of the computer is available for other work. Lockouts are provided as shown, to prevent the program from interfering with a transfer once it has begun. Transfers of data may take place to or from any part of the store, and are of the general form: transfer Y words starting at X. All transfers are fully checked, with built-in check-sums on mag- netic tape, and checked for both reading and punching of cards. Simultaneous reading and writing and program operation TIME Reading from Magnetic Tape Deck 'A'

Core store in use for a transfer

Writing to Magnetic Tape Deck 'B'

Program working independently -+ Transfer does not and only cause hesitation 'hesitating' when here because if needs the core program is not store at the same referring to time as it is core store at needed for this time a transfer I ' 100 microseconds

Lockout operation

Working Store

Lockoutf area of prevents program store reserved referring to device to program till transfer ended How ORION uses tlme

The diagram opposite illustrates the way the time-sharing system operates in a simple case. It shows a few milliseconds in the history of two programs. In this case, two summaries are being prepared simultaneously. The first of these is a sales analysis formed from recorded quantities and values on punched cards. The second is a similar summary of costing data taken from magnetic tape. At the instant at which the period illustrated starts, the first of a series of sets of cost data is being read from magnetic tape while a sales analysis program is completing the analysis of a card read in previously. The card program has priority over the tape program in order to ensure that the comparatively slow card reader is not kept waiting. On the other hand, while the sales card is being read in, the computer is able to use the time by getting on with the costing work. In general, probably four programs will be found to be the most usual combination to run at once, although provision is made for many more programs. Computer Programs Peripheral Devices @SALES @COSTS @CARD READER @TAPE DECK

Sales Analysis First cost data read from tape. ------I------When finished I4 attempt is made Program con tin ues to interrupt till next card needed ------) Sales card sales program, I ------reading sfarts but the latter Sales Program fries to refer I has higher priority I to card before it I is available I I so switch -) to Cost program I which in turn I calls for second I I cosf dafa and I then processes ------i------1Second cost first cost data. I data read from When tape 4------I ------magnetic tape transfer ended I I looks to see if I sales card is I I available. If is not. I So calls for third I

until eventu~lly I I fhe sales card is I read and this I I causes a swifch to I the sales V 00 on program ------Card reading I ends with the Sales analysis

TIME How ORION decides what to do next

In order to enable ORION to make full use of the time available during a data transfer, a complete automatic priority scanning faci- lity is provided. This enables a number of distinct programs to share the computer. As soon as any program tries to refer to data or equipment which is locked out during a data transfer, or as soon as any transfer is completed, the program priorities are automatically scanned and a switch made to an appropriate program. An essential feature of the ORION time-sharing system is the built-in program protection facility. By this means, each program has a section of the system reserved to it, consisting of two areas in the working store, an area of the drum store and a set of peri- pheral devices. Reservations are noted automatically during input of the program and the machine is so constructed that a program cannot alter or refer to the contents of any part of the system outside its own reservations. Priority scanning is carried out by a permanent program at the end of the store called the Time Sharer routine. This routine also keeps account of the computer time used by each program. When ORION reaches the end of a problem, or a batch of data (such as a pack of cards in a reader), it will not stop as previous would have done, but will enter another permanent program called the Organisation Program. This will carry out appro- priate changes in the program priority list and continue with the remaining programs. Or it may read in another program and alter the reservations accordingly. The Time Sharer is a fixed routine entered if the computer encounters a lockout and whenever a transfer comes to an end

rvore wnrcn device was responsrore fnr fhe lnrk orif (if nnv)

Record time used by thG mowam just left r 7 Enter Has the program overrun organisation its allotted duration ? program

Enter old program

I /me about ,,l,,c "bout 300 microseconds 500 microseconds The form of lnstructlons

The working store of ORION consists of 4,096 words of magnetic cores, extendable to 16,384 words. This is backed by the drum store, consisting of a number of drum units, each of which has either one or two drums. Each drum has a capacity of 16,384 words. The instruction code has been carefully chosen so as to provide a compact and rapid means of coding complex operations, such as those encountered in sorting and file handling. Most instructions are available in either a two-address or a three- address form. In the three-address form, the third address designates the destination of the result of an operation on the first two addresses, and is restricted to the first 64 words in the part of the store allocated to those programs concerned. These 64 words are known as the program's accumulators. In the two-address form, either or both the addresses will be "modified" by the contents of an accumulator. A new feature in the ORION code is that described as replacement. This enables either the first or second address in any instruction to be replaced by the contents of the register specified by the address in the original instruction, before the instruction is obeyed. This form of instruction is particularly useful when referring to items whose position in the store is random, but which can be referred to indi- rectly through an index. Addresses may be both replaced and modified, replacement happening first. For simple arithmetical instructions, the times depend on the type of instruction and the extent of replacement, as follows :- 3-address instruction -64 microseconds Unmodified 2-address instruction -48 microseconds Modified 2-address instruction - 68 microseconds and for each replaced address a further 16 microseconds. Multiplication (cumulative, double-length product) takes from 156 to 172 microseconds. Other forms of multiplication are available. The time for division depends on the binary structure of the numbers operated on. Division (e.g. unrounded with quotient and remainder) takes from 564 microseconds minimum to 1,112 micro- seeonds maximum. The statistical expectation of the time for random dividends (double-length) and divisors is 574 microseconds. Special instructions are provided for packing and unpacking fields, for radix conversion and for table search. S F TX number of bits 1 7

SIGNAL if zero, jump to organisation program FUNCTION e.g. addition X-ADDRESS is the first address, ranging over the whole store Y-ADDRESS is the second address, ranging over the whole store is either the final address in a three-address instruction, or the address of a modifier in a modified two-address instruction. There are 64 Z-addresses per program TYPE OF INSTRUCTION If both these bits are zero the instruction1 is of three-address type, e.g. "add the confents of regisfers 1000 and 2000 together and place the result in register 200." In two-address instructions these bits signify whether the X or Y addresses, or both of them, are to be modified by the addition of the contents of Z. Thus the following operafion can be carried out by a single two-address instrucfion: "add the contents of register 1000+m to the contents of register 2000+m where m is contained in register 200:' REPLACEMENT To facilitate the use of indexed or computed addresses either X or Y (or both of them) may optionally refer not to the operands, but the registers containing the addresses of the operands. Thus the following operafion may be carried out in a single instruction: "add the contents of register afm to the confents of register b+m where a, b, m are contained in registers 1000,2000,200." Advantages of using ORBON

The ORION System is supplied by the most experienced computer manufacturer in Europe.

The System is supported by first rate systems analysis, programming and maintenance engineering groups, with a proved record of achievement in putting computers to work economically and quickly.

The System has been designed by people with a background of operating experience which has influenced its conception at every stage.

The Time Sharing System particularly has been designed to make the system easy to use, from both the programming and the operating points of view.

Advanced automatic programming facilities are available, greatly simplifying the task of putting the computer to work.

Close co-operation is maintained with all manufacturers of peri- pheral equipment, which (like the computer itself) can be added to at any stage.

The Computer Centre helps in the preparation of basic file data, which can be prepared using either Pegasus or ORION equipment and kept up to date on magnetic tape, ready for use when ORION is installed. Prospective owners and users of Ferranti Computers are encouraged to gain concrete experience of computer use at an early stage by undertaking service work planned and programmed by themselves. Some Ferranti computer customers

Ferranti Computers are being used in all types of organisations, for all kinds of work, including banking, budgetary control, insurance, sales analysis, sales forecasting, production planning, material requirement calculations, costing, payroll and a wide range of scientific applications. Some of the customers for Ferranti Computers and Data-Processing Systems are given below.

Commercial Users in the U.K. British Petroleum Co. Ltd. Imperial Chemical Industries Ltd. (Dyestuffs Division). Martins Bank Ltd. The London and Manchester Assurance Co. Ltd. The Prudential Assurance Co. Ltd. The Scottish Widows' Fund and Life Assurance Society. The Standard Life Assurance Co.

General Users in the U.K. Associated Electrical Industries (Rugby) Ltd. Babcock and Wilcox Ltd. Bruce Peebles and Co. Ltd. Imperial Chemical Industries Ltd. (Central Instrument Laboratory). Shell International Petroleum Co. Ltd. The General Electric Co. Ltd. The Steel Company of Wales Ltd. The Ltd.

Commercial Users Overseas AB Datacentralen, Stockholm (Insurance). AB Turitz and Co., Gothenburg (Retail Stores). Skandia Insurance Company, Stockholm. South African Mutual Life Assurance Society, Cape Town. Svenska Flygmotor AB, Trollhattan (Production Planning).

Universities and Sckntifrc Org~nhtions The Universities of Buenos Aires, Durham, Leeds, London, Manchester, Oxford, Southampton, Stuttgart. Admiralty Research Laboratory. Air Ministry, The Meteorological Office. Council for European Nuclear Research, Geneva. D.S.I.R., The Road Research Laboratory. The Belgian Atomic Energy Authority. The Swedish Atomic Energy Authority. The United Kingdom Atomic Energy Authority, Hanvell, Risley, and Winfrith Heath.

Aircraft Design Groups British Aircraft Corporation (3 Ferranti computers) (excluding data-processing) Group (5 Ferranti computers) Royal Aircraft Establishment (2 Ferranti computers) Ferranti range of computer systems

merseus I

sirius orlon

Enquiries to FERRANTI LTD LONDON COMPUTER CENTRE 68-71 NEWMAN STREET LONDON W1 Telephone MUSeum 5040

and 21 PORTLAND PLACE LONDON W1

Oftice, Works and Research Laboratories WEST GORTON MANCHESTER 12 Telephone EASt 1301

Research Laborafories LILY HILL BERKS

LIST DC.41A November 1960